Advance in robot navigation over rough terrain using multimodal LiDAR and camera data

This paper describes a way to advance in robot navigation by development of algorithms for building Bird's-Eye View maps using fused data from a frontal cam-era and a LiDAR and usage this maps for route planning. Camera data is used to con-struct a Bird's-Eye View map via Inverse Perspective Mapping and to mark obstacles recognized by computer vision methods. LiDAR data, in turn, is used to build elevation maps of the terrain both along and across the intended path. Subsequently, the A* path planning algorithm is enhanced by incorporating an integrated cost function. This function estimates the traversability cost of a path segment by considering the terrain slope, derived from the LiDAR-based maps. This approach enables the generation of a more optimal motion profile for rough terrain navigation compared to using camera data alone.